Outline Introduction to AI ECE457 Applied Artificial Intelligence Fall 2007 Lecture #1 What is an AI? Russell & Norvig, chapter 1 Agents s Russell & Norvig, chapter 2 ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 2 Artificial Intelligence Artificial intelligence is all around us Computer players in video games Robotics Assembly-line robots, auto-pilot, Mars exploration robots, RoboCup, etc. Expert systems Medical diagnostics, business advice, technical help, etc. Natural language Spam filtering, translation, document summarization, etc. ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 3 What is an AI? Systems that Humanly Think Neural networks Act ELIZA Rationally Theorem proving Deep Blue Rationality vs. Humans: emotions, instincts, etc. Thinking vs. acting: Turing test vs. Searle s Chinese room Engineers (and this course) focus mostly on rational systems ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 4 1
Act Rationally Perceive the environment, and act so as to achieve one s goal Not necessary to do the best action There s not always an absolutely best action There s not always time to find the best action An action that s good enough can be acceptable Example: Game playing Sample approach: Tree-searching strategies Problem: Choosing what to do given the constraints Think Rationally Uses logic to reach a decision or goal via logical inferences Example: Theorem proving Sample approach: First-order logic Problems: Informal knowledge Uncertainty Search space ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 5 ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 6 Think Rationally Act Humanly 1. X = Y/Z XZ = Y 2. X = Y X + Z = Y + Z 3. X * Y + X * Z X * (Y + Z) 4. b/c = AH/b 5. a/c = BH/a 6. AH + BH = c a. b² = AH * c b. a² = BH * c c. a² + b² = BH * c + AH * c d. a² + b² = c * (AH + BH) e. a² + b² = c² Turing-test AI Improve human-machine interactions up to human-human level Drawbacks: In some cases, requires dumbing down the AI Lots of man-made devices work well because they don t imitate nature ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 7 ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 8 2
Think Humanly Cognitive science Neural networks Helps in other fields Computer vision Natural language processing Rational Agents An agent has to perceive its environment to act upon its environment A rational agent has an agent program that allows it to do the right action given its precepts Agent Program ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 9 ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 10 Types of Agents Simple reflex agent Selects action based only on current perception of the environment Model-based agent Keeps track of perception history Goal-based agent Considers what will happen given its actions Utility-based agent Adds the ability to choose between conflicting/uncertain goals Learning agent Adds the ability to learn from its experiences Simple Reflex Agent If-then Rules ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 11 ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 12 3
Simple Reflex Agent Dune II (1992) units were simple reflex agents Harvester rules: IF at refinery AND not empty THEN empty IF at refinery AND empty THEN go harvest IF harvesting AND not full THEN continue harvesting IF harvesting AND full THEN go to refinery IF under attack by infantry THEN squash them Model-Based Agent If-then Rules ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 13 ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 14 Goal-Based Agent Utility-Based Agent if I do action X if I do action X Happiness in that state Goal Utility ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 15 ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 16 4
Critic Learning Agent Feedback Performance standard Knowledge Learning Element Changes Learning Goals Performance Element Problem Generator ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 17 Properties of the Fully observable vs. partially observable See everything vs. hidden information Chess vs. Stratego Deterministic vs. stochastic vs. strategic Controlled by agent vs. randomness vs. multiagents Sudoku vs. Yahtzee vs. chess Episodic vs. sequential Independent episodes vs. series of events Face recognition vs. chess ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 18 Properties of the Static vs. dynamic vs. semi-dynamic World waits for agent vs. world goes on without agent vs. world waits but agent timed Translation vs. driving vs. chess with timer Discrete vs. continuous Finite distinct states vs. uninterrupted sequence Chess vs. driving Single agent vs. cooperative vs. competitive Alone vs. team-mates vs. opponents Sudoku vs. sport team vs. chess ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 19 Properties of the Crossword Puzzle Fully observable, deterministic, sequential, static, discrete, single-agent Monopoly Fully observable, stochastic, sequential, static, discrete, competitive multi-agent Driving a car Partially observable, stochastic, sequential, dynamic, continuous, cooperative multi-agent Assembly-line inspection robot Fully observable, deterministic, episodic, dynamic, continuous, single-agent ECE457 Applied Artificial Intelligence R. Khoury (2007) Page 20 5